AI model learns Standard Model physics directly from LHC collision data

Researchers have developed a transformer-based generative model called ShellFlow that can learn the structure of the Standard Model of particle physics directly from data collected at the Large Hadron Collider. This model, trained on approximately one billion collision events from the ATLAS Open Data release, can reproduce various physics phenomena without explicit prior knowledge beyond basic kinematic formulas. ShellFlow successfully learns intra-particle kinematics, dilepton resonances, the Weinberg angle, and the masses of the W and top quarks, demonstrating that significant portions of the Standard Model can be inferred directly from experimental data. AI

IMPACT Demonstrates AI's potential to accelerate scientific discovery by inferring complex physical laws from raw experimental data.

RANK_REASON Academic paper detailing a new method for learning physics models from experimental data. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• ATLAS Open Data
• J/psi meson
• Large Hadron Collider
• Riemannian flow matching
• Standard Model
• top quark
• Weinberg angle

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →